[MITgcm-devel] [MITgcm-cvs] MITgcm/tools/build_options CVS Commit

[David Ferreira]

On 4/27/10 3:27 AM, Menemenlis, Dimitris (3248) wrote:
> I have reverted to -mp
> sorry for careless change
> Dimitris
>    
Thanks.

Also, I've run a test report with intel-comp.11.1.046 (intead of 
intel-comp.10.1.013)
with "-fp-model precise -ftz", and unfortunately the restart is still 
broken. So there is
no escape in a new ifort version (although weirdly the testreport is a 
bit better with
ifort 11 than 10).

Also I noted that "-ftz" makes a big improvement to the testreport 
(apart from breaking the
restart), maybe the magic combination is "-mp"+"-ftz"+ifort 11...
david





> On Apr 26, 2010, at 10:00 PM, David Ferreira wrote:
>
>    
>> Dimitri,
>>      
>>>> Has anyone actually tested that "-fp-model precise -ftz" breaks the restart?
>>>>
>>>>          
>>    Looks like you've missed my e-mail:  "-fp-model precise -ftz" does
>> break the restart.
>> So I think we should get back to the old version of the optfile.
>>      
>>>> If noone has tested restarts with "-fp-model precise -ftz", I can put it on my todo list.
>>>>
>>>>          
>> If you really want to do this kind of test, I suggest doing it with a
>> newer version of ifort
>> than intel-comp.10.1.013 (because on pleiades with ifort.11, there is no
>> need for "-mp").
>> So hopefully the problem disappears by itself with higher versions ....
>> david
>>      
>
> _______________________________________________
> MITgcm-devel mailing list
> MITgcm-devel at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-devel
>